Slashdot Mirror
NAND Flash Better Than DRAM For PC Performance
Lucas123 writes "Adding NAND flash memory to a PC does more for performance than DRAM and costs less, according to a new study. As the price difference between the two memory types widens, NAND flash will become the memory of choice in the PC. The effects of NAND flash adoption are already being felt in the DRAM market, as revenue in 2011 is expected to decline 11.8%."
NAND flash degrades over time and has a limited amount of program/erase cycles.
Last time I checked, DRAM was still an order of magnitude faster than NAND flash, so swapping out your memory for flash storage would seem to be insane to me. At first I thought it was going to be how replacing a spinning disk with a flash drive is a much more noticeable upgrade than going from 4GB to 8GB of memory, but the article seems to suggest that because the market dipped a bit, DRAM is going to die out entirely and we'll be using only NAND flash for all memory on the system.
There would have to be some tremendous breakthroughs in speed, power, and especially reliability before I ever considered such a thing. It would be complete lunacy with today's technology.
I read the internet for the articles.
The article gives zero useful information and a link where you can buy the actual study. What was the pricing used for the comparison of $1 dram versus $1 nand? Surely this is OS dependent as well.
"An appropriate balance of NAND, DRAM, and an HDD yields superior performance per dollar to a simple DRAM/HDD system,"
Basically, TFA is saying that it will be awhile before we go back to a unified cache that's both RAM and storage (like core memory). Need more RAM, shrink the drive partition. Need more file storage, sacrifice RAM. It all sounds good in theory, but bus speeds and CPU technology change rapidly. I seriously doubt they can create a standardized I/O bus for removable NAND based storage devices and still keep up with future performance demands.
Life is not for the lazy.
No, but they're great in iPads.
There's no -1 for "I don't get it."
The former is not the cause of the latter. The rise of mobile devices with less DRAM in them is more likely to blame: less people are buying new PCs and Laptops when their phones and/or tablets can do everything they need.
This seems to have nothing to do with topic. Am I completely whooshed, or is this a trick to get YouTube hits?
Holy confusion. From reading the slashdot post you would think that NAND is faster than DDR3...Neither post or article are explaining this in simple terms: if you replace/augment your HDD with SDD you get more performance boost that any other upgrade would for $
I think all they mean is that dram isn't really all that cost effective as a data cache. For data that one intends to export out the network. Storing that data on a SSD, assuming it's a relatively static data set (which most is), uses far less power and costs less than purchasing an equivalent amount of DRAM (and the much larger mobo required to hold that DRAM). The access times are plenty fast enough to still saturate the network. That's all. Not rocket science.
This has been known for several years. Replicate a small server with 8-16G of ram + a 160G SSD + a 2TB HDD sits right on the sweet spot. In fact, even 4G of ram would probably be fine. The idea is not to replace your hard drive but instead to insert another layer of cheap caching to avoid having to maintain a complex, expensive, power hungry HDD storage system just to get better throughput.
-Matt
I suspect that what they're talking about is the effect of caching data from your disk drives in Flash instead of DRAM, and also letting you swap data out of DRAM into Flash instead of disk. Flash is cheap enough that for typical applications, you can cache most of your active data there, not having to wait for rotating machinery.
Windows 7 is supposed to have some feature that manages this in an intelligent way - so you can speed up your machine for a year or so by adding a $10-20 memory stick. (I'm not running Win7, so I haven't tried it - but my laptop has an SD card slot, which would let me leave a card in there full time, without it sticking out like a USB stick.)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Is the assertion really hybrid hard disks and turbo memory are having a noticable negative effect on the DRAM market?
Trying to imply a relationship between two markets by realitive growth is especially rediculous considering explosion of the smart phone market which relies entirely on flash.
The only thing more rediculous about TFA is the idea NAND is in any way a suitable replacement for DRAM.
I see this article as being myopically focused upon "main memory in portable end-user devices."
DRAM is going to stay vital for at least the server market, and I would guess the desktop market too (for as long as desktops last). Your iPad 3, maybe they have a point, but server apps would work current NAND into an early grave. The cost savings would be greatly offset by the service outages.
And since "the Cloud" is the new big thing, that means that DRAM is going to be around for a while. I don't see how the marginalization of the desktop is going to also kill the server market. The death of the PC? I'll buy that. But it's not going to kill DRAM because too many other platforms use it.
Now, if we ignore all those servers and just look at the magical end-user devices, well, that would be totally daft. Those devices aren't worth spit without something storing and delivering content to them. That something is an actual computer, rather than just a device, and it can't be using NAND for its main memory.
IMO, anyone who is doing actual computing is going to be using DRAM for the foreseeable future.
I've been telling people this ever since SSD drives came out. A system with 2GB of DRAM and an SSD drive will easily outperform a system with 8GB of DRAM with a traditional Hard Drive in every benchmark that matters to the average user. It'll boot far, far faster, programs will load instantly, defrag's are a thing of the past, virus scans take mere seconds instead of hours, and by the time your SSD drive is used up, you probably need a new computer anyways
Unless you need to use 8GB of RAM to complete your work. At which point your totally screwed. I'm pretty sure using an MLC SSD as virtual memory at this scale voids your warranty.
Now mix two Corsair SSD drives in RAID 0 like i've done for the past year along with 4GB of DRAM and the PC absolutely screams, there is no comparison, none whatsoever between traditional hard drives and SSD drives. Even (6) 15k RPM SCSI drives in stripe RAID can't keep up with the I/O of 1 SSD.
A pair of mirrored 7200 RPM disk drives with 32 GB of RAM and a warm cache runs circles around your setup. It would also be more reliable, cheaper and provide 10x the storage capacity.
Why should the controller fail, especially, why should the controller fail before the flash?
Computer memory is just fancy paper, CPUs just fancy pens with fancy erasers; the 'net is just a fancy backyard fence.
I'm more inclined to think the author has just learned what RAM is and doesn't yet understand the difference between flash RAM and dRAM.
Computer memory is just fancy paper, CPUs just fancy pens with fancy erasers; the 'net is just a fancy backyard fence.
If SSDs appear and DRAM goes down it could aso be that there are now more subnotebooks or ultraportibles in which DRAM is a power consumer and HDs are too big? It could be that MS, under the pressure of the first netbook wave which contained linux has shown reason and put out Windows 7 in opions which allow to run it on normal machines. I mean. Just thinking.
I personally dont see Flash replacing DRAM soon. I see that DRAM memories stop to grow for other reasons.
Let me say it that way round: i see that my PCs DRAM in 2000 was 128 times more than in 1990 and i see that until 2010 it has only grown by 16 times, nevertheless if the machine contains SSD or HDD.
It would be a couple GB of flash that is accessible to the OS as a block device. Then, let the OS use it for paging and for caching of frequently access files or blocks.
There should be a branding campaign so that consumers know that it is extra memory that will speed up their machine. Call it something like "turbo memory."
A pair of mirrored 7200 RPM disk drives with 32 GB of RAM and a warm cache runs circles around your setup. It would also be more reliable, cheaper and provide 10x the storage capacity.
Depends entirely on the workload. The HD drives will lose if:
a) The working set exceeds 32GB. If the accesses are random, they will lose by 2-3 orders of magnitude.
b) Whenever there is a need to commit or sync the writes to disk. On a normal system, that will happen every few seconds. If you disable that, your reliability argument is gone.
I'm assuming MS is keeping an SSD version of ReadyBoost as a Windows 8 "new feature". This should offer a very good price/performance ratio.
The Cloud - because you don't care if your apps and data are up in the air.
It is far, FAR more important for your computer to have enough RAM than to replace a HDD with an SSD. At this point (and probably for a long time) flash is not replacing DRAM. You need to have RAM in your system for it to work. Flash replaces hard disks.
Well cool, HDDs are by far the slowest component these days. SSDs are have somewhere in the range of 2-5x the transfer rate they do and more importantly are an order of magnitude or more faster on access.
Well that still is no comparison to DRAM. DDR3 is 40x the transfer rate of even fast SSDs and about 4-5 orders of magnitude less access time. So you can't just have flash, at least not if you want a nice n' fast CPU.
Now in terms of practical usage I find RAM is way, WAY more important. If you don't have enough, some programs will just flat out not run. If your system is starved, paging kills the performance, even with an SSD handling the paging. Knocking in a good amount of RAM is the #1 thing you can do to keep your system running well and it is damn cheap.
SSDs improve responsiveness, don't get me wrong. I love mine and I'm happy to have them (though to be fair I wasn't willing to get them until I saw some on sale for $200 for 256GB). However it is a more minor improvement than having a system with plenty of RAM or a good CPU. I do notice some slowness to my non-SSD work system, but not much.
The other problem is even though flash is cheaper per GB ($2ish per GB as opposed to more like $9ish for DRAM) you need more disk space than memory. My laptop has what I consider a reasonable amount of both, that is 4GB of RAM and 256GB of SSD. My desktop has a ton of RAM, 16GB, and a moderate amount of SSD, 512GB. So the SSDs cost me a hell of a lot more, despite their lower per unit cost. I could easily recommend a 4GB or more RAM upgrade to anyone, I couldn't recommend an SSD big enough to hold a good amount of stuff.
Pretty much I only recommend SSDs if you've already maxed out your RAM. Spend your money on that first, then if you are still willing to bear the cost of an SSD, go ahead.
In that vein, I noticed more improvement on my laptop than on my desktop. No small part of that is likely the RAM. The desktop has RAM to spare, it can cache a ton of stuff. The laptop is not starved for RAM, but not does it have a massive surplus. The base usage on the system is about 1.5GB for OS and background services. Gives it maybe 2.5GB for caching when nothing else is running. Hence the SSD helps more.
The right balance of NAND, DRAM and an HDD yields better results than DRAM and HDDs, study finds
So.. they have NAND, DRAM and HDDs, and they choose to kill.. DRAM? What? If something is going to fade away it is the HDDs..
They give no information about how they measured and came up with this: a dollar's worth of NAND flash improves PC performance more than adding a dollar's worth of DRAM. The closest they come to explaining it is this: After reviewing a "wide range of DRAM and NAND configurations," as well as nearly 300 industry-standard PC benchmarks
Total garbage. After working with systems that have huge amounts of RAM in them, I can only conclude they are basing this off of Microsoft's paging algorithms. Put the swap/pagefile on a NAND device and things will really speed up due to the abusively aggressive nature of Microsoft's paging algorithms. If you could make a virtual ram drive to put the page file in, it would be even faster than a NAND solution... but these guys probably do not know how to do that.
Again, if such grandiose claims are going to be made, provide some hard data.
strike
"Someone needs to talk to the tree of liberty about its ghoulish drinking problem." by ohnocitizen
A quick web search doesn't bring up any examples, but I remember when flash first came out, some of the manufacturers insisted on calling it "flash RAM" in their ads.
I also remember the arguments on the BBSes, and the conclusion that EEPROM was already different enough from UVEPROM and other ROM as to call into question the ROM part of the acronym, and that "flash ROM" seemed a bit like an oymoron. Enough people have complained about calling it flash RAM that the maufacturers have gone to "flash memory". But you still see lots of examples of the term "flash RAM" in use on the web, and it's really not technically incorrect. (Any re-writable store used to be called RAM in some camps, although you're probably too young to remember that. Sure, some other camps insisted on SRWM or the like for serially re-writable devices.)
Sure, the term causes confusion, see the article we seem to be commenting on.
Computer memory is just fancy paper, CPUs just fancy pens with fancy erasers; the 'net is just a fancy backyard fence.
ROM means Read Only Memory. Flash definitely isn't ROM in that sense: it is EEPROM, Electrically Erasable Programmable Read-Only Memory. The most common use of this before the SSD craze was the storage of BIOS settings. RAM means Random Access Memory, which typically is writeable (but doesn't have to be) as ROM usually is also RAM.
RAM and ROM are as such not mutually exclusive and mean different things. They don't even have to be silicon. Take a CD-ROM, which is read-only memory, but also random access (it's a block device when used as data. Music CDs are not random access) and they are not chips made from silicon.
Ahhh...the great dumpster continuum. Many a free computer will be found there. -- sowth (748135)
I think that's an admission that modern software is too bloated and virtual memory on a computer kills performance.
Now in terms of practical usage I find RAM is way, WAY more important. If you don't have enough, some programs will just flat out not run. If your system is starved, paging kills the performance, even with an SSD handling the paging.
I think the idea is that with the faster access times of an SSD, the operating system will be able to evict things from the system's cache more quickly, leaving more RAM for applications to use.
about 1.5GB for OS and background services.
This is part of the problem. Handheld devices are still RAM-starved, which is why handheld devices run a different incompatible operating system. But small laptops are also RAM-starved, and they run a desktop OS for the sake of user familiarity despite having only 1 GB (source: Dell.com).
I'm coming in a bit late to this debate, but I got a completely different take on what the article was about than just about every post here.
Everyone here seems to have used the article as an excuse to drag out the old "tired" debate of NAND vs HDD (and DRAM in this case) and which is better (from a technical perspective). I didn't see anything in the article that compared these on a technical front. The article was about the NAND market vs the DRAM market going into the future. It makes total sense to me.
For instance, my main development machine is (a fairly typical) quad-core machine with 8GB RAM and lots of HDD. With the various development tools I use, I generally run about 75% RAM utilization with little or no swapping. If I were to add another 8GB RAM I would see virtually no performance increase. However, if I were to add an SDD to the machine (especially if used for the OS and tools I frequently use - mostly read only files) I would see a dramatic improvement. Ergo, my money would be better spent on SDD than DRAM. That's it. That's what I took away from the article.
While one can certainly point up specific examples of where adding DRAM would increase performance more than adding NAND, that's not the point of the article.
I can certainly see the industry getting to the point where the DRAM begins to level off (how much RAM do "typical" users really need and at what point does adding DRAM stop adding value and begin to degrade the "overall" performance characteristics of the machine when power, especially battery life, is factored into the equation).
The article is all about the blend of DRAM, NAND and HDD storage and how the market for NAND is expected to rise much more rapidly than the market for DRAM, perhaps even leading to a decline in DRAM revenues.
Nowhere is the article suggesting one can directly replace DRAM with NAND, That's just silly. They are completely different technologies intended for completely different application.
Sometimes the light at the end of the tunnel is the headlight of an oncoming train.
Reading the article before commenting is not compulsory here.
"Adding NAND flash memory to a PC does more for performance than DRAM and costs less" isn't a statement about markets. It's a statement about how the use of certain technologies affect the overall performance of a PC.
The Tao of math: The numbers you can count are not the real numbers.
What about Intel's SSD lineup? (the MLC ones that is - because everything you have said only applies to MLC). They seem to have put all their focus on the controller and management aspect of it. I've been using their second generation one in my main computer for about a year now, and they seem pretty solid in my opinion.
I don't get what you are saying about power failure though... i mean i'm sure that is possible with some poorly designed controller/firmware. but surely most of them are smart enough to design one where in the case of power failure during page remapping for wear leveling ; ether A) mapping storage design restricts data loss to the pages being operated on B) their is a power buffer large enough to allow the controller to finish internal operations, or C) both.I'm pretty sure any of the major SSDs intended for main block device replacement implement something along those lines.
We use raw flash, with linux and JFFS2 or UBIFS. Which is a filesystem designed to run on flash, raw flash. Wake me when "SSD" offers that solution. (TRIM is not a solution, it's a workaround)
For non esoteric circumstances, it seems quite committing to adopt a file-system that specifically caters for MLC NAND's wear and granularity limitations. As great as it is for bringing large fast SSDs to us for cheap right now... it certainly wont be around for long, there are probably going to be lots of different types of solid state storage technologies competing for the place of block storage device, can you imagine having a different file-system for every single one?.
Another problem with offloading MLC page management to the file-system is that you are also then committing to retaining wear leveling data and page pools. it just completely stops it being an easy swappable wipeable block device... IMO leave the management specific to the storage technology on the technology, keep the file-system as relevant to data / OS as possible
Also if your going to be purist enough to complain about TRIM being a work around then you might as well also call SATA a work around... the next level is bootable pcie, wakeup... everything is a workaround, if it doesn't work with current technology it doesn't sell, if it doesn't sell it wont get developed.
"Adding NAND flash memory to a PC does more for performance than DRAM and costs less" isn't a statement about markets. It's a statement about how the use of certain technologies affect the overall performance of a PC.
But it doesn't say that in the article. That's what the summary says. Everyone who's been here a while knows that the summary is usually way off the mark as to the true content of the story.
Sometimes the light at the end of the tunnel is the headlight of an oncoming train.
(is FAT32 still used for these high >4GB densities?)
AFAIK, Windows does not like NTFS on a "removable" drive, so unless you make the CF card appear as a hard drive (which would be OK for Windows), you have to use FAT or FAT32.
Why ? The extra RAM could be used as additional disk buffers. The fact that your current utilization is only 75% doesn't mean that extra RAM won't be used. It could be that the OS has a strategy to keep some free memory around to start new applications.
For example, my 4GB machine currently has 1.6GB free, which is an even lower utilization that yours.
(is FAT32 still used for these high >4GB densities?)
AFAIK, Windows does not like NTFS on a "removable" drive, so unless you make the CF card appear as a hard drive (which would be OK for Windows), you have to use FAT or FAT32.
AFAIK you're wrong.
I've formatted plenty of external hard drives / flash drives as FAT32, NTFS, and exFAT under Windows.
Well, an external hard drive is detected as a "hard drive", so it is partitioned and formatted like an internal hard drive. USB flash sticks and various cards in card readers are detected as "removable disk" and Widows does not want to create more than one partition and format it in NTFS.
Well, at least Windows XP does this.
The greatest performance limiter of today's top-of-the-line computers is the mechanical hard disk. Although the article may have some concepts screwed up (I assume), the one upgrade that makes the greatest _apparent_ improvement in performance is switching from an HDD to an SSD. Yes, it's true that more DRAM helps, but any decent NEW system today already comes with 4GB of RAM (2GB if you really cheap out). 4GB is overkill for most users, so switching to an SSD will have a greater impact on general performance than upgrading to 8GB of RAM. This is because 4GB of RAM is more than enough cache their most frequently used disk blocks, so the only performance hit people will see is the time to load a program that isn't already cached in RAM. Everything else will _already_ be in RAM.
Now, I'm a power user, and going from 4GB to 8GB makes a huge difference for me. I always have several apps open at once and maybe 50 tabs in Safari. (Also Mac OS X has a shameful VM compard to Linux, so the extra RAM helps even more on a Mac. Linux seems to do a way better job of figuring out what to keep cached and paged in when under memory pressure.) But as it is, because everything else is so fast (Sandy Bridge, all the RAM, etc.), the only thing that limits peformance for me is the mechanical hard drive. It's very noticable, especially after I've rebooted. My problem is that I wanted the larger drive and didn't want to shell out $1500 for the 512GB SSD. I had to make a compromise because I'm still a grad student.
I read the same article. The problem is that the entire article should be rendered as fine print underneath a HOLLYWOOD sized YMMV billboard.
Also the TCO on SSD taints the economics: failure modes are changing at a speed that gives even a hard core geek a rational ignorance skin rash. There was a nice post above on the physics of desperation. I can add DRAM without a major update to my mental catalog of high-performance electronics failure modes.
Also I don't get the economic argument in the first place. With 8GB of RAM, I rarely find myself disk bound except on huge datasets that also verge on CPU bound.
I don't think they are talking about a speed increase of the kind where you could integrate over a 24 hour period and declare "you saved 10 minutes today". They are talking about speed which shows up in a higher derivative, making "attention span" an essential parameter of the workload simulated.
HDD performance is within *my* attention span 99% of the time. Or put it another way, every application I use regularly is permanently open on one desktop or another. Starting an application fresh every 15 minutes and manually placing the newly open windows where I wish them to be is *not* within my attention span.
Does this incredible (and cheap) SSD performance increment save more time than the average person wastes in a day shuffling window placement on applications of short duration?
I bet not.
You're wrong about that, you can format any removable device as NTFS if it is large enough. I do it with USB flash drives all the time and have done it to sd cards in the past.
I don't agree, mostly because the abbreviations have meaning and you totally disregard those. It might be that it is "generally accepted", but it puts your RAM and ROM definition in the same line as a "PIN Number" or and "ATM Machine": the abbreviation has become void of all meaning.
Ahhh...the great dumpster continuum. Many a free computer will be found there. -- sowth (748135)
It is by my definition, and I said so in my original post.
Ahhh...the great dumpster continuum. Many a free computer will be found there. -- sowth (748135)
Depends entirely on the workload. The HD drives will lose if:
a) The working set exceeds 32GB. If the accesses are random, they will lose by 2-3 orders of magnitude.
If the working set exceeds 32GB would you expect the system with only 4GB of memory to have a chance? I'm sure you could think about it for a while and dream up a mythical workload with just the right characteristics to make it work.. for a few hours anyway until all the oxide on the flash gates is gone.
8GB ram = $50.
b) Whenever there is a need to commit or sync the writes to disk. On a normal system, that will happen every few seconds. If you disable that, your reliability argument is gone
I was thinking more in terms of storage reliability. When SSDs fail they more often fail specatucularly with no warning. RAID0 is less reliable than single disk. A failure of any disk means the entire array is unusable.
In terms of flush I would never advocate disabling writes to persistant storage under any circumstance. A performant system would coalesce as many operations into a transaction as reasonable to minimize the need for excessive synchronization.